Mihaela Simionescu ISSN 2071-789X RECENT ISSUES IN ECONOMIC DEVELOPMENT Economics & Sociology, Vol. 7, No 1, 2014 48 Mihaela Simionescu PhD, Senior Researcher Institute for Economic Forecasting TESTING SIGMA CONVERGENCE ACROSS EU-28 of the Romanian Academy Bucharest, Romania E-mail: [email protected]ABSTRACT. The main goal of this research is to assess the degree of convergence in European Union- 28 (EU- 28) during 2000-2012. After a spatial analysis of macroeconomic indicators using maps and graphs that were built in GeoDa, the sigma convergence was tested. For this purpose, variation indicators in the simple and weighted variant for GDP per capita in PPS were utilized, the weights being given by the population weights in each country of the EU-28. The results indicated a decrease in divergence process in 2012 compared to 2000, but there is still not enough evidence for the closeness of an acceptable degree of convergence. However, there is a low degree of divergence in EU-28 compared to EU-27 in 2012 with respect to 2000. In 2012, the EU-28 populations explained better the GDP compared to the number of employed persons. Received: December, 2013 1st Revision: March, 2014 Accepted: April, 2014 DOI: 10.14254/2071- 789X.2014/7-1/5 JEL Classification: C18, C88, F2, F15 Keywords: convergence, sigma convergence, GDP per capita, standard deviation, variance, coefficient of variation. Introduction For measuring the degree of realization for the convergence process one should assess the levels of different indicators that refer to: variability/homogeneity, polarization, concentration, complementarity, entropy. These indicators might confirm or not different aspects of the convergence process. The most known and applied is the sigma convergence measured by variation indicators. This paper is structured in several parts. After this brief introduction, a short literature review is made, underlying the latest results regarding the convergence assessment. The methodological part consists in the presentation of the statistical indicators used in measuring the variation in GDP/capita. The empirical application supposes the calculation of variation measures for European Union-28. The results put into evidence that during 2000-2012 the convergence criterion was not fulfilled, even if the degree of variability decreased in time. A section dedicated to main conclusions was presented in the end. 1. Literature According to Sala-i-Martin (1996), and indicators are new tools for measuring the degree of convergence and the speed for getting convergence. indicator shows the convergence and divergence tendency depending on the value of sample variance. Mihaela Simionescu, Testing Sigma Convergence Across EU-28, Economics & Sociology, Vol. 7, No 1, 2014, pp. 48-60. DOI: 10.14254/2071-789X.2014/7- 1/5
13
Embed
Mihaela Simionescu, Testing Sigma Convergence Across EU-28, … · 2016-09-29 · degree of convergence and the speed for getting convergence. indicator shows the convergence and
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Mihaela Simionescu ISSN 2071-789X
RECENT ISSUES IN ECONOMIC DEVELOPMENT
Economics & Sociology, Vol. 7, No 1, 2014
48
Mihaela Simionescu PhD, Senior Researcher Institute for Economic Forecasting
TESTING SIGMA CONVERGENCE ACROSS EU-28
of the Romanian Academy Bucharest, Romania E-mail: [email protected]
ABSTRACT. The main goal of this research is to assess the degree of convergence in European Union- 28 (EU-28) during 2000-2012. After a spatial analysis of macroeconomic indicators using maps and graphs that were built in GeoDa, the sigma convergence was tested. For this purpose, variation indicators in the simple and weighted variant for GDP per capita in PPS were utilized, the weights being given by the population weights in each country of the EU-28. The results indicated a decrease in divergence process in 2012 compared to 2000, but there is still not enough evidence for the closeness of an acceptable degree of convergence. However, there is a low degree of divergence in EU-28 compared to EU-27 in 2012 with respect to 2000. In 2012, the EU-28 populations explained better the GDP compared to the number of employed persons.
parameter shows the speed for accomplish the convergence when it has a negative sign.
Mankiw, Romer and Weil (1992) and Islam (1995) introduced in the model control variables
like saving rate and population growth, showing that the economies with a low initial income
tend to grow faster than the economies with higher income. On the other hand, alternative
econometric models were developed by Quah (1996) and Durlauf (1996) to show that the
transversal growth model is consistent with the multiple mechanisms of endogenous growth
and inconsistent with the convergence. The behavior and evolution of convergence-clubs
were deeply analyzed. The conventional theory of convergence and the empirical researches
hide the convergence-clubs and the polarization of countries in rich in poor. Some economists
like Friendman (1992) and Quah (1996) considered the beta indicator as irrelevant for the real
convergence process of the economic growth. There are several forms for this tool utilized in
econometric analysis: absolute beta convergence, club convergence and conditional beta
convergence. The beta and sigma indicators are related and reciprocal checked. The
assumption of diminishing yield of the neo-classical theory states that poor economies tend to
grow faster than rich ones. This conclusion has the following implications: the coefficient of
variation for GDP/capita slowly decreases and there is an inverse relationship between the
rate of economic growth for GDP/capita and the initial level of GDP/capita. However, the Sala-i-Martin (1996) suggested that different relationships may occur between these two types
of convergence.
Azomahou, El ouardighi, Nguyen-Van, and Cuong Pham (2011) used a semi-
parametric partially linear model to evaluate the convergence across EU members, indicating
no convergence for high income countries. Beyaert and García-Solanes (2014) evaluated the
effect of economic conditions on long-term economic convergence. The GDP/capita
convergence is not the same on the business cycle during 1953-2010. Cuaresma, Havettová
and Lábaj (2013) evaluated the income convergence dynamics using forecast models for
European countries, the future human capital investment being a factor that will generate
income convergence.
Palan and Schmiedeberg (2010) tested the structural convergence for Western
European countries using unemployment rate data. The divergence is obvious in technology-
intensive manufacturing industries. Le Pen (2011) applied the pair-wise approach of
convergence proposed by Pesaran (2007) to the output per capita of some European regions.
Crespo-Cuaresma and Fernández-Amador (2013) evaluated the dynamics of the
standard deviation for European area business cycles to figure out the convergence patterns.
In eighties a relevant business cycle divergence was observed and in nineties a persistent
convergence was assessed.
Kutan and Yigit (2009) used a panel data analysis for 8 new members of the European
Union and showed that human capital is the most important factor that determined
productivity growth during 1995-2006. Monfort, Cuestas, and Ordóñez (2013) used a cluster
analysis, observing two convergence clubs in EU-14 countries, the Eastern European
members being also divided in two groups.
Iancu (2009) used the sigma convergence to measure the evolution of real
convergence process between the EU countries on three groups: EU-25, EU-15 and EU-10,
the coefficient of variation indicating a divergence growth of the economies during 1995-
2006. Mihuţ and Luţaş (2013) measured the sigma convergences across the new members of
the European Union.
2. Methodology
The methodological approach regarding the convergence is based on the scientific
contribution brought by Solow. This neoclassical model had a huge impact on the new models
Mihaela Simionescu ISSN 2071-789X
RECENT ISSUES IN ECONOMIC DEVELOPMENT
Economics & Sociology, Vol. 7, No 1, 2014
50
in order to diminish the gap between the real conditions of the economy and the different
versions. The economic growth could be supported by investments that generate positive
externalities associated with the development of human capital. In the new approach of
convergence, the human and physical capital together with the technologic process
contributed to the convergence realization. On the other hand, the econometric methods are
more used in testing the assumptions of the models.
The variation is measured for more elements like regions and countries using simple
indicators (the deviation, the range) and synthetic indicators (average linear deviation, variance,
standard deviation, and coefficient of variation). The synthetic indicators express the distance
between a variable values for each of the element and the average level. The convergence stage
at a certain moment is assessed using an indicator of variation that shows rather the opposite
aspect of convergence, because it expresses in a numerical form how far the elements of the‘
entire are from the central level towards the values of the indicator converges.
In a dynamic analysis the variation in decrease allows us to conclude the existence of
a more obvious convergence process. The most useful indicator is the coefficient of variation,
because it allows making comparisons. The variance is computed as:
𝜎2 =1
𝑛 (𝑦𝑖 − 𝑦 )2𝑛
𝑖=1 , (1)
𝑦𝑖 – the variable, i – index for spatial elements (regions, countries), 𝑦 – simple arithmetic
average (𝑦 =1
𝑛 𝑦𝑖
𝑛𝑖=1 ).
The variance expresses the degree of variation of the values compared to the average.
It is affected by outliers and by the variable measurement of unit. For small samples the
denominator is replaced with (n-1) instead of n. The variance is used to calculate the standard
deviation (𝜎 = 𝜎2) and the coefficient of variation (𝐶𝑉 =𝜎
𝑦 ), the last one expressing in a
relative form the variation compared to average. According to Villaverde Castro (2004), the
coefficient of variation is indicated in convergence analysis because it does not depend on the
unit of measurement and the measure order of the indicators.
A decrease in time of the values for standard deviation indicates the fulfillment of
convergence from the point of view of y variable. This aspect is known as convergence. It
is recommended the use of coefficient of variation based on the population weight:
𝐶𝑉 ′ = 𝜎 ′2
𝑦 , (2)
where 𝜎 ′2 =1
𝑛 𝑦𝑖 − 𝑦 2𝑛
𝑖=1 (𝑝𝑖 − 1), 𝑝𝑖 – the weight of region/country population.
Analytical tools and indicators that show diminish of differences compared to average
or diminish of differences between two or more time series are used to measure the
convergence:
lim𝑡→∞ 𝑥 − 𝑦 = 𝑐, (3)
The stochastic or non-stochastic principle might be used to assess the decrease of
differences.
The indicator called sigma convergence is used to characterize the level of
convergence by measuring the variance of GDP per capita for one year, utilizing the cross-
section data about regions and countries. The indicator is relevant when comparisons are
made. For describing the convergence tendency, time series are used on a discrete interval
Mihaela Simionescu ISSN 2071-789X
RECENT ISSUES IN ECONOMIC DEVELOPMENT
Economics & Sociology, Vol. 7, No 1, 2014
51
from t to t+T. In a certain time period when the variance of the variable decrease (the
indicator value decrease in time), the convergence process took place: 𝜎𝑡+𝑇 < 𝜎𝑡 . When the
variance grows the divergence process took place: 𝜎𝑡+𝑇 > 𝜎𝑡 .
The theoretical assumption that should be checked using data of a representative
sample is:
1
𝑇log
𝑦𝑖′ 𝑡0+𝑇
𝑦𝑖′ 𝑡0
= 𝑐 − (1−𝑒−𝛽𝑇
𝑇) log 𝑦𝑖 ,𝑡0 + 𝜀𝑖 ′ 𝑡0 ,𝑡0+𝑇, (4)
The catch up of the rich economies is confirmed by the decrease of the GDP/capita
degree of variation between countries, but also by the negative sign of the annual convergence
rate of the GDP/capita of countries in the sample, these countries arriving at the same time in
the steady state.
3. The sigma convergence in European Union
The first part of this empirical application is dedicated to a spatial analysis of
macroeconomic indicators in the entire Europe that includes EU-28 and only in EU-28.
The descriptive analysis is done for having a picture of the macroeconomic and
population state in the entire Europe. For the GDP (thousand Euro) in Europe a map was
created in GeoDa software to visualize the four groups of countries according to the GDP
level (thousand Euro) in 2012.
The four groups of countries in both maps are denoted by 1 from 4, the value 4
corresponding to the group of countries with the darkest nuance of orange.
Group 1 – [0; 8538]
Group 2 – [8578; 3.908e+004]
Group 3 – [4.494e+004; 2.726e+005]
Group 4 – [2.896e+005; 2.356e+006]
Figure 1. Groups of countries in Europe according to GDP level in thousand Euro in 2012
Source: author’s graph in GeoDa.
Mihaela Simionescu ISSN 2071-789X
RECENT ISSUES IN ECONOMIC DEVELOPMENT
Economics & Sociology, Vol. 7, No 1, 2014
52
Group1 – [932; 1.337+e006]
Group 2 – [2.04e+006; 5.354e+006]
Group 3 – [5.419e+006; 1.08e+007]
Group 4 – [1.132e+007; 1.029e+008]
Figure 2. Groups of countries in Europe according to population in 2012
Source: author’s graph in GeoDa.
The box maps are utilized to determine the outliers in a stricter way than the percentile
maps. The values are grouped in 6 categories: two outlier categories and the quartiles: 1-25%,
25-50%, 50-75%, and 75-100%. The outliers are determined using the following rule: the
outliers are the values 1.5 times greater than the difference between the third percentile and
the first percentile.
Figure 3. Box map for GDP (thousand Euro) in EU-28 in 2012
Source: author’s graph in GeoDa.
Mihaela Simionescu ISSN 2071-789X
RECENT ISSUES IN ECONOMIC DEVELOPMENT
Economics & Sociology, Vol. 7, No 1, 2014
53
The box map for GDP (thousand Euro) shows that there is a quite high range between
values. There are 5 countries with higher values than the quartiles.
A cartogram is a way to show the extreme values on a map. In this study a circular
cartogram was realized in GeoDa, the countries being represented by circles. The area of each
circle is proportional to the value of the chosen variable. In his case, the selected variable was
GDP/capita. A nonlinear optimization routine was applied in order to place the circles close to
the initial location of the matching countries. The cartogram is useful to link the actual spatial
layout and the idealized layout of the observations.
Figure 4. The cartogram of the European countries (50) according to GDP/capita in PPS
Source: author’s graph in GeoDa.
In this case the hinge is 1.5. Because of the nonlinear optimization procedure, the
cartogram could be improved. In this cased more 20 000 iterations were used and the circles
seem to jump.
Figure 5. Improved cartogram for 20 000 more iterations of the European countries (50)
according to GDP/capita in PPS
Source: author’s graph in GeoDa.
The parallel coordinate plot (PCP) is a good alternative to box map, the main
advantage of PCP being the identification of observations that cluster in a multivariate space.
The variables are seen in a multivariate comparison and they become parallel axes, on an axis
the values being represented from the minimum value to the maximum value. Line segments
Mihaela Simionescu ISSN 2071-789X
RECENT ISSUES IN ECONOMIC DEVELOPMENT
Economics & Sociology, Vol. 7, No 1, 2014
54
are used to represent an observation with more variables. These line segments are the
counterpart of the point in a multidimensional scatter plot.
Each variable was rescaled, the lowest value being on the left and the highest one on
the right. The observations are ordered and located relative to the range of the values.
Figure 6. Parallel coordinate plot for GDP (thousand Euro) and Population in EU-28 (2012)
Source: author’s graph in GeoDa.
Figure 7. Parallel coordinate plot for GDP and Population growth in EU-28 (2012)
Source: author’s graph in GeoDa.
There are quite different PCPs if we chose GDP (thousand Euro) and population on a
hand, and GDP (thousand Euro) and population growth on the other hand. Germany is the
country with the highest GDP and with the largest population. The second PCP indicates that
indeed Germany has the highest value of GDP (thousand Euro), but Ireland has the greatest
population growth in 2012 compared to 2011.
The regression analysis was applied to determine the factor that explained better the
GDP (thousand Euro) of EU-28 in 2012. Simple linear regressions were proposed, the
explanatory variables being the population and the number of employed persons. The
parameters of the regression models were estimated in GeoDa using ordinary least squares
method and the assumptions (the auto-correlation, the homoscedasticity and the normality of
errors) were checked using the proper tests. The models were valid, because the t-computed
for the slope is higher than the critical value at a 5% significance level. Moreover, the p-
values associated to the slopes are less than 0.05. The following two graphs indicate a positive
Mihaela Simionescu ISSN 2071-789X
RECENT ISSUES IN ECONOMIC DEVELOPMENT
Economics & Sociology, Vol. 7, No 1, 2014
55
relationship between GDP (thousand Euro) and population, but also between GDP (thousand
Euro) and employed persons.
According to Jarque-Bera test, the assumption of errors’ normality is not rejected, the
JB statistic being less than the critical value (5.991). However, all the applied tests for
heteroscedasticity (White test, Breusch-Pagan test and Koenker-Bassett test) indicated the
same results: the errors are not homoscedastic. The results of the parameters’ estimations
were presented in Appendix 1 for the first model and Appendix 2 for the second regression
model.
Figure 8. The relationship between GDP (thousand Euro) and population in EU-28 (2012)
Source: author’s graph in GeoDa.
Figure 9. The relationship between GDP (thousand Euro) and employed persons in EU-28
(2012)
Source: author’s graph in GeoDa.
Mihaela Simionescu ISSN 2071-789X
RECENT ISSUES IN ECONOMIC DEVELOPMENT
Economics & Sociology, Vol. 7, No 1, 2014
56
The two models were compared according to Akaike criterion. The differences
between the values of Akaike for the two models are insignificant. However, a slightly lower
value was obtained for the first model. This indicates that the populations of EU-28 countries
explained better the GDP (thousand Euro) in 2012 compared to employed people.
The data referring to GDP/capita in PPS EU27=100 (volume indices of real
expenditure per capita in PPS EU27=100) for 2000 and 2012 are provided by Eurostat. The
dataset was expressed as ratios relative to the EU average in 2000 and 2012. The variance and
the standard deviation were computed for EU-27 and EU-28 in 2000 and 2012. The results of
the computations are presented in Table 1.
Table 1. Measures of variability for GDP/capita in PPS (EU27=100) in EU27 and EU28
Region Year Variance Standard
deviation
Coefficient of
variation (%)
EU-28 2000 0.2342 0.4839 48.39%
EU-28 2012 0.1766 0.4203 42.03%
EU-27 2000 0.2203 0.4694 46.94%
EU-27 2012 0.1746 0.4178 41.78%
Source: author’s computations.
The reason for making the comparison between EU-27 and EU-28 is related to the fact
that the previous studies did not take into account the convergence in EU-28 and we might be
interested to see that the adding of a country generates a progress in convergence process. It is
important here to check if the new member of EU had a good contribution to the realization of
convergence process or a negative influence by the accentuation of the divergence.
In both years (2000 and 2012) the dataset for GDP/capita in PPS EU27 is
heterogeneous, the coefficient of variation being greater than the threshold of 30-35%.
However, in 2012 in both regions (EU27 and EU28) the coefficient of variation decreased
from 46.94% to 41.78%, respectively from 48.39% to 42.03%. Therefore, the conclusion
would be that in the analyzed period the degree of variation decreased and even if the
economic convergence is not confirmed by the variability measures some improvements were
made in this direction.
The other method with EU27 and EU28 populations’ weights is used. The weights are
determined as the population of each country over the overall population of the EU in 2000,
respectively 2012.
Table 2. Measures of variability for GDP/capita in PPS (EU27=100) in EU27 and EU28 using
populations’ weights
Region Year Variance Standard
deviation
Coefficient of
variation (%)
EU-28 2000 0.2134 0.4619 44.45%
EU-28 2012 0.1684 0.4104 40.35%
EU-27 2000 0.2178 0.4666 44.53%
EU-27 2012 0.1662 0.4077 40.37%
Source: author’s computations.
The variability measures indicate that there was an obvious decrease of the indicators
in 2012 compared to 2000 in both regions (EU27 and EU28). This implies that the weighted
Mihaela Simionescu ISSN 2071-789X
RECENT ISSUES IN ECONOMIC DEVELOPMENT
Economics & Sociology, Vol. 7, No 1, 2014
57
method suggested an improvement in the achievement of this convergence criterion.
However, the indicators’ values still show a slow degree of convergence. In EU28 the
coefficient of variation is lower than the value registered for EU27, but the difference is
almost insignificant. The coefficient of variation could be used to make comparisons with the
results of the previous method. Lower values were obtained for the weighted method that
indicates a better degree of convergence in EU.
Table 3. Variability indicators for real expenditure in PPS of EU28 (purchasing power parity
in EU28) during 2000-2012
Region Variance Standard
deviation
Coefficient of
variation
EU-28 0.001111 0.033337 3.3356%
EU-27 0.001114 0.033378 3.3397%
Source: author’s computations.
The results in dynamics for real expenditure in PPS indicate that during 2000-2012 a
slow degree of variability in GDP per capita in PPS for EU28 purchasing power parity was
registered with a minor decrease for EU-28 compared to EU-27.
Conclusions
The sigma convergence was assessed in this study for European Union using statistical
variation measures. Therefore, standard deviation and coefficient of variation for GDP/capita
in PPS were computed in two versions: the standard version and the variant that takes into
account population’s weight.
During 2000-2012 the degree of variation decreased, but the economic convergence is
not confirmed, the coefficient of variation being more than 40% for both methods. However,
the weighted method suggested an improvement in the achievement of this convergence
criterion
The analysis in dynamics of the real expenditure in PPS showed that during 2000-
2012 a slow degree of variability in GDP per capita in PPS for EU28 purchasing power was
observed.
All in all, it is important to measure the achievement of the sigma criterion of
convergence in time to get the proper decisions for improve the degree of convergence. This
study for EU-28 is relevant, by taking into account the situation of the new EU member-
Croatia. The research might be developed by assessing the other indicators that could be a
good hint for showing the degree of convergence process.
References
Albu, L. L. & Roudoi, A. (2003), Scenarios Of Economic Development In Romania –
Medium To Long-Term Forecasting Models, Journal for Economic Forecasting,
Institute for Economic Forecasting, Vol. 0(5), December, pp. 64-77.